PaperlayerView exact revision on arXiv

Source-linked AI summary

Soft robotic suits: State of the art, core technologies and open challenges

Michele Xiloyannis, Ryan Alicea, Anna-Maria Georgarakis, Florian L. Haufe, Peter Wolf, Lorenzo Masia, Robert Riener

arXiv:2105.10588v2cs.ROeess.SY

TL;DR

Soft robotic suits address the limitations of rigid exoskeletons by using compliant wearable structures to assist human movement. This paper defines and classifies these systems, reviews their actuation, interfaces, and intention detection, and finds reported benefits for locomotion and motor impairment support while identifying adaptability and hardware limitations.

  • Problem

    The paper addresses how soft wearable robots can provide assistance while reducing the restriction, complexity, and accessibility problems associated with rigid exoskeletons.

  • Method

    The authors critically review compliant actuation, physical human-suit interfaces, intention-detection strategies, and biomechanical effects across soft robotic suits.

  • Results

    Soft robotic suits have reduced metabolic cost during walking and running and improved gait-related outcomes in people with mobility impairments.

  • Takeaways & Limitations

    The review highlights soft robotic suits as a developing approach for augmenting human movement and supporting rehabilitation across locomotion and upper-limb tasks.

  • Takeaways & Limitations

    Existing lower-limb intention-detection controllers have limited adaptability to changing speeds, slopes, carried loads, and tasks such as stair climbing or jumping.

Abstract

from arXiv · show

Wearable robots are undergoing a disruptive transition, from the rigid machines that populated the science-fiction world in the early eighties to lightweight robotic apparel, hardly distinguishable from our daily clothes. In less than a decade of development, soft robotic suits have achieved important results in human motor assistance and augmentation. In this paper, we start by giving a definition of soft robotic suits and proposing a taxonomy to classify existing systems. We then critically review the modes of actuation, the physical human-robot interface and the intention-detection strategies of state of the art soft robotic suits, highlighting the advantages and limitations of different approaches. Finally, we discuss the impact of this new technology on human movements, for both augmenting human function and supporting motor impairments, and identify areas that are in need of further development.

I. INTRODUCTION

Soft robotic suits emerged as lighter, more wearable alternatives to rigid exoskeletons, addressing the complexity, weight, cost, and biomechanical restrictions of rigid frames. The paper defines these devices, classifies them, and critically reviews their technologies and open challenges.

  • Introduction: Rigid exoskeleton complexity increases cost, weight, and size, while joint misalignment can generate uncontrolled interaction forces.
  • Introduction: Rigid exoskeletons evolved from metal, ground-supported frames toward textile- and elastomer-based wearable robots that better conform to human anatomy.The Hardiman weighed 348 kg, whereas the HAL and Harvard exosuit weighed 15 kg and 5 kg, respectively.
  • Scope: The review focuses on lower-limb, trunk, and upper-limb assistance, while intentionally omitting hand-function technologies and avoiding a comprehensive device list.
  • Definition: Soft robotic suits are clothing-like devices that work in parallel with muscles and use the human body, rather than a rigid frame, to transfer reaction forces.They therefore function as an external layer of muscles rather than an external skeleton.
  • Taxonomy: The paper classifies soft robotic suits by power source into passive, passive-adaptive, and externally powered devices using tensile or expansive functional units.

III. SOFT ROBOTIC SUITS FOR THE LOWER LIMBS

Lower-limb soft robotic suits have progressed from early tendon concepts to textile systems using tensile actuation, with designs targeting propulsion, joint assistance, rehabilitation, and reduced metabolic cost. Their core advantages include low impedance, portability, and the ability to become mechanically transparent.

  • A. Tensile robotic suits: Kelly’s 1919 Pedomotor used artificial tendons parallel to the leg muscles and anticipated backpack-based timing of assistance.
  • A. Tensile robotic suits: Harvard’s 2013 exosuits used textile frames and short, well-timed assistance pulses to reduce walking metabolic cost without imposing kinematic restrictions.
  • A. Tensile robotic suits: Tensile lower-limb suits use McKibben pneumatic artificial muscles, electric motor-tendon units, twisted-string actuators, and multi-articular tendon paths.
  • A. Tensile robotic suits: A key advantage is that the suit can transition from applying high forces to becoming mechanically transparent, while Bowden cables allow proximal actuator placement near the wearer’s center of mass.
  • A. Tensile robotic suits: Lower-limb designs target hip, knee, and ankle assistance across walking, running, mobility impairment, and rehabilitation applications.

B. Expansive robotic suits

Expansive robotic suits use inflatable Pneumatic Interference Actuators to generate joint moments while preserving movement freedom. Demonstrated systems assist ankle plantar-flexion and knee extension during walking.

  • B. Expansive robotic suits: The ExoBoot uses an inflatable Pneumatic Interference Actuator to assist ankle plantar-flexion without restricting other ankle degrees of freedom.
  • B. Expansive robotic suits: A PIA-based suit assists knee extension during swing, using inert fabric containment, IMU gait detection, and closed-loop pressure control.The actuator delivered approximately 25% of the biological torque required for knee extension.
  • B. Expansive robotic suits: Expansive actuation can be combined with tensile pneumatic muscles to provide distinct assistance phases at the ankle.

C. Passive-adaptive robotic suits

Passive-adaptive and passive suits regulate or recycle mechanical energy rather than directly supplying continuous external power. The XoSoft modulates elastic assistance with clutches, while the ExoBand stores and returns walking energy using only passive elements.

  • C. Passive-adaptive robotic suits: XoSoft uses electromagnetic clutches to engage and disengage passive elastic elements during the gait cycle, controlling when energy is stored and released.
  • D. Passive robotic suits: The ExoBand weighs 645 g and stores energy before maximum hip extension, returning it during swing to support limb advancement.
  • D. Passive robotic suits: At 1.1 m/s, the ExoBand reduced walking metabolic cost by an average 3.3% versus unassisted walking in nine older adults.

A. Tensile robotic suits

Tensile suits transmit assistive forces by tensioning contractile elements arranged alongside human muscles. The reviewed systems span pneumatic muscles, electric motor-tendon units, shape-memory alloys, and hybrid designs for upper-limb and trunk assistance.

  • Examples: McKibben pneumatic muscles were used in early shoulder-assistance suits and later arranged in parallel or series to trade higher force for longer stroke.The miniature-actuator network used parallel arrangements for higher forces and series arrangements for longer strokes.
  • Examples: Electric motor-tendon suits supported feeding, elbow motion, industrial lifting, and multi-joint assistance for users with muscular weakness or stroke.Reported systems combined off-board or mobile motor-tendon units with passive or soft components, including the CRUX suit’s distributed soft-rigid load paths.
  • Design space: Upper-limb tensile suits use diverse architectures, including artificial tendons, pneumatic muscles, elastic elements, and combinations of soft and rigid components.The figure includes shoulder, elbow, feeding, lumbar, and multi-joint assistance systems.
  • Examples: A shape-memory-alloy elbow suit generated peak forces of 120 N with a 24 g actuator, illustrating the high force-to-weight potential of tensile actuation.The actuator used Nickel-Titanium coil springs arranged in parallel.

B. Expansive robotic suits

Expansive suits generate assistance through pressurized structures that deform and interact with the body. The section also situates these systems within broader actuation, interface, and control developments for wearable assistance.

  • Pneumatic assistance: Inflatable PIAs can support shoulder abduction, stabilize the arm against gravity, and assist elbow flexion through bending moments generated by bladder deformation or interference.A Y-shaped PIA delivered up to 16 Nm−1 at 90° abduction and 136 kPa, while an adjacent-chamber array produced elbow assistance.
  • Tendon-driven systems: Tendon-driven systems benefit from remote actuation, established motor-control technology, and the ability to slacken into passive garments during selected gait phases.Bowden cable routing enables flexible transmission between distant points, but transmission friction and compliance reduce efficiency.
  • Tendon-driven systems: Tendon-driven suits impose high skin shear and low mechanical efficiency, while friction, backlash, and stick-slip complicate accurate force tracking.These effects can require larger power supplies and degrade accuracy, bandwidth, and stability.
  • Control: Force-tracking approaches include force-based position control and admittance-based cascaded velocity-force control, whereas impedance control remained unimplemented in the reviewed soft suits.The survey attributes the latter gap partly to instability when rendering stiff force behavior.

2) PAMs:

Pneumatic artificial muscles generate tensile force by inflating an elastic bladder inside an inextensible braid. They offer muscle-like compliance and high power-to-weight ratio, but compressors, nonlinearities, and limited control flexibility constrain wearable deployment.

  • Operating principle: McKibben PAMs use an elastic airtight bladder within helically woven inextensible fibers; inflation expands the bladder radially and contracts it longitudinally.This converts pneumatic pressure into tensile mechanical power.
  • Advantages: PAMs provide high power-to-weight ratio, muscle-like force-length behavior, intrinsic compliance, and strokes up to 25% of resting length.Their decreasing force with increasing contraction length also contributes to intrinsic safety.
  • Limitations: Heavy and bulky compressors or compressed-air sources limit PAM portability, motivating research into more energy-dense pneumatic sources such as hydrogen peroxide reactions.The proposed chemical approach is described as a promising path rather than an established solution.
  • Comparison: Compared with electric motor-tendon units, PAMs and related pneumatic systems face more difficult portability and control trade-offs despite their compliant actuation.The figure contrasts tensile actuation mechanisms, including electric motor-tendon units, PAMs, twisted strings, shape-memory alloys, and PIAs.
  • Control: PAM nonlinearities have led many suits to use practical controllers such as bang-bang, feedforward position, proportional feedback position, or PID pressure regulation.These approaches are effective but provide less fine tuning of assistive profiles than electric motor-tendon actuation.

4) Shape-memory alloys:

Shape-memory alloys contract when heated and return toward their memorized shape during cooling. Their high force density and low profile suit wearable actuation, but low efficiency and slow relaxation constrain practical use.

  • Operating principle: Shape-memory alloy wires, commonly Nickel-Titanium, contract under Joule heating and relax toward their memorized shape when cooled.The actuation mechanism is thermally driven rather than pneumatically or electromagnetically driven.
  • Capabilities: A fabric muscle combining 20 SMA coiled wires produced up to 120 N contraction force, a 67% stroke, and low mass.The cited actuator demonstrates the high force-to-weight potential of SMA-based suit actuation.
  • Limitations: SMA actuation is constrained by efficiency below 3% in typical cases and relaxation times that can reach 30 s without active cooling.Forced air or liquid immersion can accelerate cooling by up to two orders of magnitude but is impractical for wearable applications.
  • Control: SMA suits use position or temperature feedback, including PI control with wire encoders and temperature control with thermocouples.The reviewed implementations regulate either contraction length or actuator temperature.
  • Field-level perspective: Across soft suits, electric motor-tendon units produced the most promising effects on human motor performance because motors and their electronics are mature and readily controllable.The survey also notes that emerging smart textiles and other soft actuation and sensing methods may change future suit designs.

B. Physical human-suit interface

The physical human-suit interface must transmit actuator forces efficiently while remaining comfortable and compatible with human movement. Key design challenges include pressure and shear management, power losses, force-transfer delay, and the trade-off between stiffness and comfort.

  • Interface principles: Soft robotic suits transfer actuator forces through fabric and elastomer load paths rather than a rigid frame, making textile engineering central to pHRI.Their defining interface challenge is creating effective load paths from actuators to the human skeleton.
  • Comfort and pressure: Pressure distribution depends more on padding stiffness than thickness, while interface geometry and conformable fabrics can improve attachment-point design.Pressure-monitoring studies identified peaks at attachment points and informed data-driven optimization of interface topology and materials.
  • Interface principles: Efficient pHRI anchors suits to stiff body locations, uses stiff materials and direct force paths, minimizes shear, and maximizes suit–skin contact area.The Harvard exosuit guidelines associate the first three principles with power transfer and the latter two with comfort.
  • Comfort and pressure: Circumferential compression becomes uncomfortable at approximately 16-34 kPa and painful at approximately 20-27 kPa, with lower pain thresholds in people with chronic pain.These findings motivate stricter pressure requirements for healthcare applications.
  • Comfort and pressure: Tangential forces affect comfort and blood flow, yet methods for measuring and reducing shear at textile-skin interfaces remain scarce.Recent work has proposed simultaneous measurement of shear and normal forces between a textile cuff and skin.
  • Power transfer and comfort: Soft tissues, fabrics, and flexible transmissions can absorb up to 55% of actuator power, while compliance delays force transfer and can reduce assistance effectiveness.A stiff transmission improves mechanical efficiency but often reduces comfort, creating a direct design trade-off.
  • Advantages and open challenges: Soft materials avoid rigid-exoskeleton misalignment and can become dynamically transparent when assistance is disengaged, but full-day use still requires improved comfort.Textile anchor points and inflatable actuators can have negligible mass relative to human limbs.

C. Sensors and strategies for intention-detection

Soft-suit intention detection uses mechanical manifestations of movement or neural signals to time and adapt assistance. Mechanical approaches are robust but less adaptable, while neural approaches are promising yet complex; upper-limb control remains especially dependent on manual input.

  • Sensing strategies: Intention-detection systems use mechanical signals such as load, pressure, position, and inertial measurements, or neural signals such as muscle activity.Mechanically intrinsic controllers are the predominant approach because their measurements are robust and repeatable.
  • Sensing strategies: Neural control can precede movement and adapt to environmental dynamics, but noisy signals, skin impedance, fatigue, and frequent calibration make it more complex.These factors help explain why mechanically intrinsic control remains more common in soft robotic suits.
  • Lower-limb control: Lower-limb mechanically intrinsic controllers detect gait events with foot-switches, pressure sensors, or IMUs to segment steps and trigger assistive profiles.These approaches exploit the rhythmic structure of locomotion to deliver assistance at appropriate times.
  • Open control challenges: Mechanical controllers have limited adaptability to speed, slopes, carried loads, and tasks such as jumping, stair climbing, and downhill walking.This limitation can reduce device benefit and, in the worst case, restrict movement or induce falls.
  • Lower-limb control: Mechanical controllers adapt assistance timing to walking cadence, including by updating the gait period from the average period of the last five steps.Other approaches use the delay between known events within a gait cycle for more responsive timing adaptation.
  • Lower-limb control: 100% accuracy was achieved when vertical CoM acceleration at maximum hip extension distinguished walking from running and selected the corresponding assistance profile.The acceleration and thigh-angle signals were obtained using IMUs on the abdomen and thighs.
  • Upper-limb control: Upper-limb suits often use manual on/off inputs because arm movements vary more than rhythmic walking, although this approach adds cognitive burden.A capacitive stretch sensor has also been used to monitor elbow angle for gravity compensation.

VI. EFFECTS ON HUMAN MOVEMENTS

Soft robotic suits have produced measurable metabolic and gait benefits across walking, running, and mobility-impaired populations, although outcomes depend on device and experimental conditions. Studies use net and gross metabolic changes to distinguish assistance from no-suit and powered-off references.

  • Metabolic and gait effects: Soft robotic suits have generated quantitative benefits for walking, running, and impaired gait, with outcomes reported using net and gross metabolic changes.The review notes that device characteristics and experimental conditions vary across studies, limiting direct comparison.
  • Walking: −22.8% was the peak gross metabolic change reported for human walking assistance.For loaded walking, the highest reported values were −22% gross and −14.9% net metabolic change.
  • Running: −14.6% was the highest reported net metabolic reduction for running, achieved with a rigid ankle-assistance exoskeleton using off-board actuation.A soft hip-extension device at 2.5 m/s produced −4% net and −9.1% gross metabolic changes.
  • Mobility impairments: A unilateral tethered suit improved swing dorsiflexion and forward propulsion in nine stroke survivors, producing a more symmetric gait and −10% gross metabolic change.A later multisite clinical trial found a similar commercialized device safe and feasible for post-stroke gait rehabilitation.
  • Mobility impairments: 0.16 m/s was the clinically meaningful walking-speed improvement reported for a person with incomplete spinal cord injury using the Myosuit.The same study reported a 9% reduction in transport cost versus no suit on an outdoor sloped mountain path.
  • Passive assistance: 3% was the Exoband’s reduction in net walking metabolic cost among older adults using only elastic bands for limb advancement.The result emphasizes considering passive device dynamics during wearable-device design.

B. Soft robotic suits for the upper limbs and trunk

Upper-limb and trunk suits have reduced muscle activity across shoulder, elbow, arm, and back tasks, while studies also examine rehabilitation and neuromotor adaptation. Results use heterogeneous measures, with surface-EMG activity commonly serving as the outcome.

  • Reported effects: Upper-limb and trunk studies report muscle-activity reductions across assistance for shoulder, elbow, lifting, isometric, and leaning tasks.The reviewed studies use less homogeneous outcome measures than locomotion studies, commonly including relative surface-EMG changes.
  • Shoulder assistance: −79.5% was the reported net change in anterior and medial deltoid activity for a PIA-based device assisting shoulder movements.The device also improved workspace area in chronic stroke patients.
  • Elbow assistance: Up to −73%, −77%, and −59.7% changes were reported for biceps brachii, brachioradialis, and triceps brachii activity, respectively, using an EMG-based controller.These were among the highest reductions reported around the elbow joint.
  • Lifting assistance: −48% and −32% gross reductions were achieved in biceps brachii and anterior deltoid activity during lifting with a tendon-driven soft suit.An expansive-PIA device produced a −63% gross change in biceps activity during an isometric task.
  • Passive trunk assistance: −43% was the peak net reduction in erector-spinae activity during a leaning task with a passive elastic suit.The result came from elastic bands designed to unload the erector spinae muscle group.
  • Neuromotor adaptation: Soft robotic suits provide a platform for studying how users learn assistance, the neural processes involved, and the effects of neuromotor impairments.Their mechanical transparency, simplicity, and versatility are identified as useful for these questions.

VII. GENERAL CONCLUSIONS

Soft robotic suits are moving wearable-robot testing beyond laboratories into daily and outdoor environments, but broader functional impact depends on resolving force transfer, actuation, and task-adaptation challenges.

  • Real-world deployment: Soft robotic suits have enabled testing in daily environments, rough terrains, and sloped mountain paths beyond laboratories, clinics, and hospitals.Their simplicity and portability support this expansion of testing contexts.
  • Real-world deployment: Future outdoor and domestic studies are needed to determine the technology’s functional impact on people’s lives.The review frames movement beyond laboratory confinement as necessary for answering these questions.
  • Technological challenges: The field still lacks a clear solution for attaching robots and transferring forces to the human body, an under-addressed challenge requiring systematic investigation.The review identifies this as the foremost technological challenge.
  • Technological challenges: Actuation remains largely dominated by electric motors, while PAM, PIA, and pneumatic-energy approaches still present control and energy-delivery challenges.Advances in these areas are described as likely to have a lasting impact on the field.
  • Technological challenges: Wearable robots must adapt seamlessly across diverse tasks and dynamics, including walking, stairs, jumping, sitting, running, changing loads, and movement speeds.This versatility is required as the devices enter daily environments.
Loading 2105.10588v2…